An improved method for semi-automated identification of submarine canyons and sea channels using digital bathymetric analysis

Submarine canyons are the key structures of the transitional area between the deep ocean floor and continental shelf known as continental slope. They are important submarine features with high ecological, economic, and scientific value. With continual improvements to global and regional marine bathymetry products, digital bathymetric analysis (DBA) techniques must evolve to support ongoing marine geological research and to advance international bathymetric feature catalogs and mapping products. This study aims to further advance the delineation of submarine canyons and associated sea channels using DBA and the recent open-access GEBCO₂019 global bathymetry grid with 15 arc-second resolution, for a small study region in the southern Celtic Sea. A modified semi-automated delineation method is presentedbased on combining hydrological network analysis and the topographic position index (TPI), a commonly used parameter available in conventional digital terrain analysis toolsets. For the case-study area of interest, 96 submarine canyons were identified, with higher morphological detail than the previous ETOPO1 and SRTM30PLUS based studies, which identified 81 and 52 submarine canyons, respectively for the same area of interest. The improvements include an increased number of vertices and limbs per canyon, increased positional precision for canyon starting points and endpoints, and the ability to identify associated sea channels. The results also highlight the importance of pre-processing and parameter localization to effectively exploit higher-resolution bathymetric data, which should also improve scale adaptability for application in different types of continental margins and with high-resolution bathymetry products. Future studies will benefit from the TPI-based identification process proposed by this study to identify submarine canyons on other continental slope regions or to delineate other linear depressed features.